Clinical Cancer Research Joint Metastasis Research Society-AACR Conference on Metastasis Translational Cancer Medicine 2008: Cancer Clinical Trials and Personalized Medicine
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Cancer Research Clinical Cancer Research
Cancer Epidemiology Biomarkers & Prevention Molecular Cancer Therapeutics
Molecular Cancer Research Cancer Prevention Research
Cancer Prevention Journals Portal Cancer Reviews Online
Annual Meeting Education Book Cell Growth & Differentiation
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplementary Data
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Wamunyokoli, F. W.
Right arrow Articles by Birrer, M. J.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Wamunyokoli, F. W.
Right arrow Articles by Birrer, M. J.
Clinical Cancer Research Vol. 12, 690-700, February 2006
© 2006 American Association for Cancer Research

Human Cancer Biology

Expression Profiling of Mucinous Tumors of the Ovary Identifies Genes of Clinicopathologic Importance

Fred W. Wamunyokoli1, Tomas Bonome1, Ji-Young Lee2, Colleen M. Feltmate2, William R. Welch3, Mike Radonovich1, Cindy Pise-Masison1, John Brady1, Ke Hao4, Ross S. Berkowitz2, Samuel Mok2 and Michael J. Birrer1

Authors' Affiliations: 1 Cell and Cancer Biology Branch, National Cancer Institute, NIH, Bethesda, Maryland; 2 Laboratory of Gynecological Oncology and 3 Department of Pathology, Brigham and Women's Hospital, Harvard Medical School; and 4 Department of Biostatistics, Harvard School of Public Health, Boston, Massachusetts

Requests for reprints: Michael J. Birrer, Cell and Cancer Biology Branch, National Cancer Institute, NIH, Room 1130, Building 37, 9000 Rockville Pike, Bethesda, MD 20892. Phone: 301-402-9586; Fax: 301-480-4756; E-mail: birrerm{at}mail.nih.gov.


   Abstract
Top
 Abstract
Materials and Methods
 Results
 Discussion
 References
 
Purpose: To elucidate the molecular mechanisms contributing to the unique clinicopathologic characteristics of mucinous ovarian carcinoma, global gene expression profiling of mucinous ovarian tumors was carried out.

Experimental Design: Gene expression profiling was completed for 25 microdissected mucinous tumors [6 cystadenomas, 10 low malignant potential (LMP) tumors, and 9 adenocarcinomas] using Affymetrix U133 Plus 2.0 oligonucleotide microarrays. Hierarchical clustering and binary tree prediction analysis were used to determine the relationships among mucinous specimens and a series of previously profiled microdissected serous tumors and normal ovarian surface epithelium. PathwayAssist software was used to identify putative signaling pathways involved in the development of mucinous LMP tumors and adenocarcinomas.

Results: Comparison of the gene profiles between mucinous tumors and normal ovarian epithelial cells identified 1,599, 2,916, and 1,765 differentially expressed in genes in the cystadenomas, LMP tumors, and adenocarcinomas, respectively. Hierarchical clustering showed that mucinous and serous LMP tumors are distinct. In addition, there was a close association of mucinous LMP tumors and adenocarcinomas with serous adenocarcinomas. Binary tree prediction revealed increased heterogeneity among mucinous tumors compared with their serous counterparts. Furthermore, the cystadenomas coexpressed a subset of genes that were differentially regulated in LMP and adenocarcinoma specimens compared with normal ovarian surface epithelium. PathwayAssist highlighted pathways with expression of genes involved in drug resistance in both LMP and adenocarcinoma samples. In addition, genes involved in cytoskeletal regulation were specifically up-regulated in the mucinous adenocarcinomas.

Conclusions: These data provide a useful basis for understanding the molecular events leading to the development and progression of mucinous ovarian cancer.

Epithelial ovarian cancer is the fourth leading cause of cancer deaths in women in the United States (1). An estimated 22,220 new cases of this malignancy will be diagnosed, and ~16,210 deaths attributed to this disease in the United States during 2005 (1). This high case fatality rate is due in part to the fact that the majority of patients (75%) are diagnosed after extraovarian spread of the disease has occurred. The 5-year survival rate for women with the late-stage disease is 25% compared with a rate of >90% for women with early-stage (2, 3).

The histologic classification of ovarian carcinomas is based on morphologic criteria and corresponds to the different types of epithelia in the female reproductive system, including papillary serous, mucinous, endometrioid, clear cell, and Brenner (transitional; ref. 4). Each class has been further subclassified into benign, malignant, and borderline or low malignant potential (LMP) to reflect their histopathology. Mucinous ovarian tumors account for 12% to 15% of all ovarian neoplasms. The majority of mucinous ovarian tumors are benign (75%), whereas borderline and adenocarcinomas account for 10% and 15%, respectively (4).

Although mucinous ovarian tumors represent a small subset of ovarian carcinomas, they possess distinct clinical characteristics. First, there is a high frequency of intestinal differentiation in LMP and malignant subtypes of mucinous tumors, which is also observed in metastatic gastrointestinal neoplasms. The similarity in morphology among these tumors necessitates the performance of an appendectomy and the use of specific immunostains to reliably distinguish mucinous ovarian cancers from those of gastrointestinal origin (5). Second, mucinous LMP tumors without invasive metastasis are typically treated by surgical removal alone and result in excellent survival (6, 7). Mucinous LMP tumors with invasive metastasis are frequently treated with chemotherapy after surgery although its effectiveness remains controversial (8). Advanced mucinous ovarian adenocarcinomas respond poorly to first-line platinum-based chemotherapy compared with other epithelial ovarian tumors (9). Consequently, the aggressive nature of advanced-stage mucinous ovarian cancers results in a significantly reduced survival rate for women with these tumors when compared with other histologic subtypes (10–12). Finally, there is increasing evidence, both molecular and pathologic, supporting a unique progression model for mucinous ovarian cancer. Several studies have suggested that a proportion of mucinous ovarian cancers, unlike other histologic subtypes, such as serous, progresses from a benign cyst to an LMP tumor before developing into an adenocarcinoma. Evidence to support this transitional progression paradigm comes from electron microscopy studies, histology, and the coexistence of cells with varying degrees of malignancy in mucinous ovarian cancer, including studies of mucinous carcinomas demonstrating the presence of prominent foci of benign-appearing or atypical mucinous epithelium (13, 14). In addition, transitions between benign and malignant areas are seen in 80% malignant mucinous adenocarcinomas (15). Further studies have shown similar molecular alterations in all mucinous ovarian tumor types. For example, identical K-ras mutations are frequently found in coexisting borderline and invasive epithelia within a mucinous tumor (16). More importantly, these mutations are not related to the stage of the disease in mucinous tumors, unlike nonmucinous tumors, suggesting a continuum in the development of this cancer (17–20). Given these unique attributes of mucinous ovarian tumors, there is a need to carry out further studies that will decipher and elucidate the observed morphologic, immunohistochemical, and molecular characteristics they exhibit.

Gene expression profiling has been used to identify important genes contributing to the process of ovarian tumorigenesis and delineate differences between tumors of different histology and stage (21). To further understand the unique clinicopathologic characteristics of mucinous ovarian carcinoma, we developed gene expression profiles for a set of mucinous ovarian tumors using the Affymetrix U133 Plus 2.0 oligonucleotide array (Affymetrix, Santa Clara, CA), which represents over 47,000 transcripts and variants, and compared these with profiles of papillary serous tumors. The addition of 10 normal ovarian surface epithelium (OSE) brushings facilitated tumor to normal class comparisons. Our microarray analysis suggests that mucinous LMP tumors and adenocarcinomas are most closely related to serous adenocarcinomas. Furthermore, the close association of LMP and adenocarcinomas, relative to their serous counterparts, implies that they may follow an incremental course of development culminating in advanced disease. Finally, we identified differentially expressed genes that may contribute to the distinct characteristics of mucinous tumors.


   Materials and Methods
Top
 Abstract
 Materials and Methods
Results
 Discussion
 References
 
Tissue samples
A total of 25 mucinous specimens (6 cystadenomas, 11 LMP tumors, and 8 adenocarcinomas) were analyzed in this study. In addition, a series of previously completed serous (20 LMP tumors and 20 adenocarcinomas) and 10 normal OSE arrays were included (Table 1; ref. 22). All of the tumor specimens were obtained from previously untreated ovarian cancer patients at the Brigham and Women's Hospital (Boston, MA) and verified by a gynecologic oncology pathologist (W.R. Welch) before their use. The classification of the tumor specimens was determined according to the International Federation of Gynecology and Obstetrics standards. Appendixes from all cases were removed to confirm that they had normal histopathology. Normal OSE isolates were obtained from the ovaries of patients undergoing surgery or other gynecologic diseases. All of the specimens used in this study, and their corresponding clinical information, were procured according to the Institutional Review Board protocols.


View this table:
[in this window]
[in a new window]
  		Table 1. Mucinous cystadenomas, LMP, and invasive tumor specimens included in the microarray analysis

 
Microdissection
Frozen tumor samples were sectioned (7 µm) and affixed onto FRAME slides (Leica, Germany). They were subsequently fixed (70% ethanol, 30 seconds), stained (1% methyl green), rinsed in deionized water, and air-dried. Microdissection was carried out under an MD LMD laser microdissecting microscope (Leica, Germany). Tumor cells (20,000-25,000) were collected and used in the isolation of RNA. The acquisition and cellular characterization of the OSE specimens has been described previously (23).

RNA isolation and quality analysis
Total RNA was extracted from tumor cells and OSE brushings using RNeasy Micro kit (Qiagen, Valencia, CA) according to the instructions of the manufacturer. The quality and quantity of the isolated RNA was analyzed using a Bioanalyzer 2100 system (Agilent, Palo Alto, CA).

Amplification of RNA and array hybridization
A two-cycle amplification protocol (Affymetrix) was used to generate sufficient cRNA for microarray analysis according to the recommendations of the manufacturer. Briefly, 10 ng total RNA was used in the first-round synthesis of double-stranded cDNA. The RNA was reverse transcribed using a two-cycle cDNA synthesis kit (Affymetrix) followed by amplification using a MEGAscript T7 kit (Ambion, Austin, TX). The cRNA generated from the first-round synthesis was cleaned using a GeneChip sample clean-up Module IVT column (Affymetrix). The second-round double-strand cDNA amplification was carried out using an IVT labeling kit (Affymetrix). The resulting biotin-labeled cRNA was purified using an IVT clean-up kit (Affymetrix) and quantified using a UV spectrophotometer (A260/280; Beckman, Palo Alto, CA). An aliquot (15 µg) of cRNA was fragmented by heat and ion-mediated hydrolysis at 94°C for 35 minutes. Fragmented cRNA was hybridized in a hybridization oven (16 hours, 45°C) to a human genome U133 Plus 2.0 GeneChip oligonucleotide array (Affymetrix) representing 47,000 transcripts and variants, including 38,500 well-characterized human genes. The washing and staining of the arrays with phycoerythrin-conjugated streptavidin (Molecular Probes, Eugene, OR) was completed in Fluidics Station 450 (Affymetrix). The arrays were subsequently scanned using a confocal laser GeneChip Scanner 3000 and GeneChip operating software (Affymetrix).

Microarray analysis
Data normalization. All 75 arrays were normalized at a target value of 500 using the GeneChip operating software (Affymetrix). This was followed by uploading of the data into the Microarray Analysis Database of the National Cancer Institute (http://nciarray.nci.nih.gov/index.shtml) for quality control screening and collation before further analysis. Normalized data was filtered using Biometric Research Branch ArrayTools version 3.2.2 software developed by Dr. Richard Simon and Amy Peng Lam (http://linus.nci.nih.gov/BRB-ArrayTools.html). This multifunctional Excel software is able to process and analyze data using the R version 2.0.1 environment. Hybridization control probe sets and probe sets scored either as absent at {alpha}1 = 0.05 or marginal at {alpha}2 = 0.065 were excluded from further analysis. Only those transcripts present in >50% of the arrays and displaying a variance in the top 50th percentile were evaluated further. To ensure that microdissected samples were not contaminated with infiltrating leukocytes, which might affect the quality of the data analysis, CD45 gene expression was assessed across all of the microarrays. For all 75 arrays, CD45 expression could not be reliably detected at a normalized signal intensity >100, indicating that there was no significant leukocyte contamination.

Hierarchical clustering and binary tree analysis. Hierarchical clustering was done on the filtered data set using a 1-correlation with centroid linkage in dChip version 1.3 software (24). To further determine the relationships among the different pathologic subtypes within each tumor type, binary tree analysis was done in Biometric Research Branch ArrayTools using a diagonal linear discriminant prediction method. This analysis determined the misclassification of samples within each class (subtype) providing an independent assessment of relationships among the specimens based on their expression profile.

Class comparison and pathway analysis. Differentially expressed genes were identified for tumor versus tumor and tumor versus OSE analyses using a multivariate permutation test in Biometric Research Branch ArrayTools. A total of 2,000 permutations were completed to identify the list of probe sets containing fewer than 10 false positives at a confidence of 95%. Differential expression was considered significant at P < 0.001. A random-variance t test was selected to permit the sharing information among probe sets within class variation without assuming that all of the probe sets possess the same variance. A global assessment of whether expression profiles were different between classes was also done. During each permutation, the class labels were reassigned randomly and the P values for each probe set were recalculated. The proportion of permutations yielding at least as many significant genes as the actual data set at P < 0.001 was reported as the significance level of the global test. The identification of coregulated pathways contributing to the phenotypes associated with mucinous LMP tumors and adenocarcinomas was determined using PathwayAssist version 3.0 software (Iobion Informatics LLC, La Jolla, CA).

Quantitative real-time PCR validation
Quantitative real-time PCR (qRT-PCR) was done on 100 ng of amplified cRNA from the 25 mucinous tumors and 10 OSE specimens included in the microarray analysis. Primer sets specific for 14 genes selected from the lists of differentially expressed genes identified for mucinous cystadenoma, LMP, adenocarcinoma samples were used. The housekeeping genes GAPDH, GUSB, and cyclophilin were selected for normalization. An iCycler iQ Real-Time PCR Detection System (Bio-Rad, Hercules, CA) was used in conjunction with QuantiTect SYBR Green RT-PCR kit (Qiagen, Valencia, CA) according to the instructions of the manufacturer. To calculate the relative expression for each gene, the 2–{Delta}{Delta}CT method was used using an average CT value for the three housekeeping genes (25).

Immunohistochemistry
Five-micrometer sections of 59 mucinous tumors and six normal ovarian epithelium on a tissue microarray were obtained from the Gynecologic Oncology Group tissue bank. Immunohistochemical analysis was carried out for the AGR2 protein using the ABC kit (Vector Laboratories, Burlingame, CA) according to the instructions of the manufacturer. Briefly, the tissue microarray slide was deparaffinized and rehydrated before hydrogen peroxide treatment for 30 minutes at room temperature. Antigen retrieval was done by microwave heating sections in 10 mmol/L sodium citrate buffer (pH 6) for 5 minutes. After blocking of nonspecific binding, rabbit polyclonal anti-AGR2 (kindly donated by Prof. Jin-San Zhang, Mayo Clinic College of Medicine, Rochester, MN) at a dilution of 1:100 was added and the slides incubated at 4°C overnight. Normal human lung tissue section was used as a positive control (26). Scoring of immunostained sections of normal ovarian epithelium, mucinous cystadenoma, LMP, and adenocarcinoma specimens was based on the percentage of immunopositive cells (0-100) and staining intensity (0-4).


   Results
Top
 Abstract
 Materials and Methods
 Results
Discussion
 References
 
Hierarchical clustering and binary tree analysis of mucinous and serous LMP tumors and adenocarcinomas. Genome-wide expression profiles for 19 microdissected mucinous ovarian LMP tumors and adenocarcinoma specimens were generated using the Affymetrix U133 Plus 2.0 GeneChip oligonucleotide microarray. Data filtering identified ~18,559 probe sets possessing an informative signal intensity in >50% of the arrays and a variance in the top 50th percentile. These data were compared with previously completed profiles for 40 microdissected serous tumors and 10 OSE brushings. To characterize the relationships among these serous and mucinous samples, 1-correlation metric using centroid linkage was applied to all 18,559 probe sets. A dendrogram containing two distinct arms was identified (Fig. 1A). Serous LMP tumors were most closely allied with OSE specimens (Fig. 1A, node A), whereas serous high-grade and mucinous LMP tumors associated at node B. Among the mucinous tumors, low-grade cancers were aligned with the LMP tumors in a single group (Fig. 1A, node C). High-grade mucinous adenocarcinomas formed a distinct branch at node C. Despite the existence of well-defined clusters, a minority of specimens failed to classify correctly, including two serous LMP tumors and a low-grade mucinous cancer.


Figure 1
View larger version (24K):
[in this window]
[in a new window]
  		Fig. 1. Unsupervised hierarchical clustering and binary tree analysis of OSE, mucinous LMP tumors, mucinous adenocarcinomas, and serous tumors. A, hierarchical clustering illustrated OSE specimens grouped independently of serous LMP tumors (node A), whereas high-grade (G3) serous cancers were closely associated with mucinous LMP tumors and adenocarcinomas (node B). Mucinous LMP and low-grade (G1) adenocarcinomas clustered in a branch distinct from high-grade specimens (node C). Incorrectly classified specimens are in bold italics. B, binary tree analysis resolves hierarchical clustering results. Mucinous LMP and low-grade (G1) specimens were nearly indistinguishable (node B). Mucinous high-grade (G2) adenocarcinomas were more closely aligned to LMP tumors (node C) than their serous counterparts (node A). Percentages indicate the misclassification error associated within each node.

 
To further resolve the relationships among serous and mucinous LMP tumors and adenocarcinomas, binary tree prediction using a diagonal linear discriminant predictor was applied to the 18,559 informative probe sets used in the hierarchical clustering analysis at a threshold of 0.5 (Fig. 1B). Only tumor specimens were included in the analysis to ensure the predictor would optimally resolve the relationships among the tumor classes. An estimation of the prediction error of the classifier was obtained using leave-one-out cross-validation of the entire tree building process. At node A of Fig. 1B, a misclassification rate of 7.5% was observed between serous LMP and high-grade tumors. Mucinous LMP and grades 1 and 2 adenocarcinomas were nearly indistinguishable with a misclassification rate of 40.0% (Fig. 1B, node B). High-grade mucinous tumors were classified as distinct from LMP and grade 1 adenocarcinomas (Fig. 1B, node C); however, a 21.1% misclassification rate implies that mucinous LMP and grades 1 and 2 adenocarcinomas are more closely related than their serous counterparts.

Hierarchical clustering and binary tree analysis of mucinous tumors. To ascertain the relationship among all mucinous tumors, a series of mucinous cystadenomas were included in an unsupervised analysis using 12,871 probe sets. Hierarchical clustering using 1-correlation metric with centroid linkage revealed distinct grouping (Fig. 2A) of the cystadenomas at node A, whereas the LMP and adenocarcinoma specimens retained their clustering relationships. Binary tree analysis identified a misclassification rate of 8.8% between the mucinous cystadenomas and the other tumors during binary tree analysis (Fig. 2A), suggesting that there are a number of coregulated genes shared among these groups. In addition, mucinous LMP, grade 1, and grade 2 tumors possessed misclassification rates (Fig. 2B, nodes B and C) comparable with those observed in Fig. 1B.


Figure 2
View larger version (17K):
[in this window]
[in a new window]
  		Fig. 2. Unsupervised hierarchical clustering and binary tree analysis of mucinous tumors. A, hierarchical clustering showed mucinous cystadenomas clustering independently (node A), whereas the remaining mucinous tumors were closely associated (node C). B, binary analysis revealed a small misclassification rate for mucinous cystadenomas. In contrast, LMP and adenocarcinoma samples retained a larger misclassification rates (nodes B and C). Percentages indicate the misclassification error associated with each node.

 
Identification of differentially regulated genes among mucinous tumors and normal OSE. To identify differentially expressed genes accounting for the phenotypes associated with mucinous ovarian LMP tumors and adenocarcinomas, filtering was completed for all 75 arrays to facilitate tumor versus OSE class comparisons. A total of 1,599, 2,916, and 1,765 genes were identified for mucinous cystadenoma, LMP tumors, and grades 1 and 2 adenocarcinomas, respectively, using a multivariate permutation test providing 95% confidence the number of false discoveries did not exceed 10 at P < 0.001. Each comparison returned a global significance value <0.05, indicating the number of identified probe sets would not be expected at random. A complete list of differentially regulated gene probes sets for each mucinous tumor type is detailed as Supplementary Data.

Having identified differentially expressed genes between each mucinous tumor type and normal OSE, genes encoding secretory and membrane-bound proteins, which may serve as unique markers of the ovarian mucinous phenotype, were compiled (Table 2A). In addition, we identified genes that may contribute to tumor progression or serve as effectors of this process in mucinous ovarian cancer. This was accomplished by ensuring that (a) the genes were only differentially expressed in LMP (Table 2B) or adenocarcinoma (Table 2C) specimens versus normal OSE, and (b) they remained differentially regulated when compared directly to the cystadenoma samples (a complete list is available as Supplementary Data).


View this table:
[in this window]
[in a new window]
  		Table 2. Differentially expressed genes (P < 0.001) unique to mucinous tumors versus serous specimens that are secreted or membrane bound (A), genes up-regulated in mucinous LMP tumors versus OSE and cystadenomas (B), and genes up-regulated in mucinous adenocarcinomas versus OSE and cystadenomas (C)

 
Identification of signaling pathways in mucinous ovarian cancer. PathwayAssist software was used to determine the interaction of differentially expressed genes identified in the microarray analysis and postulate how these interactions might contribute to the clinicopathologic features of mucinous ovarian cancer. Figure 3 illustrates the interactions among some of the differentially regulated genes in both mucinous LMP tumors and adenocarcinomas. The genes comprising the pathway have been implicated in the development of multidrug resistance (ABCC3 and ABCC6), signal transduction (SPRY1 and CAV-1), cytoskeleton rearrangement/signal transduction (RAC1, CDC42, RALA, IQGAP2, Cortactin), cell cycle regulation, and proliferation (CCND1, ERBB3, transforming growth factor-{alpha}), and transformation (c-JUN, K-ras2, ECT2 YES1; Fig. 3). Genes involved in the development of drug resistance were consistently up-regulated in both mucinous LMP and adenocarcinoma specimens, suggesting that development of drug resistance mechanisms in mucinous ovarian tumors is an early event. In addition, genes involved in cell cycle regulation, cellular proliferation, and transformation were also up-regulated in both LMP and carcinomas. Conversely, the majority of genes involved in cytoskeletal modulation and motility were up-regulated in adenocarcinomas. With the exception of WT1, CAV1, c-JUN, and MUC1, which were differentially regulated in all the mucinous ovarian tumors, the remaining genes in the pathway were only differentially expressed in the LMP tumors and adenocarcinomas. This suggests that a subset of genes associated with the ovarian mucinous phenotype might be involved in the progression of mucinous ovarian cancer.


Figure 3
View larger version (19K):
[in this window]
[in a new window]
  		Fig. 3. Schematic representation of putative signaling pathways identified in mucinous ovarian tumors versus OSE. Incorporating the differentially expressed genes into PathwayAssist developed a pathway. Genes included in the pathway were required to have a fold-change value ≥1.5. Multiple probe sets were averaged for each gene. Open oval, gene is up-regulated versus OSE; filled oval, down-regulation; gray oval, a gene displaying no change in expression. Genes in rectangular shape are unique to mucinous LMP, genes in hexagon are unique to mucinous adenocarcinomas, whereas genes in oval shape are common to both mucinous LMP and adenocarcinoma specimens. Table 3
Table 3. Gene annotation information

LocusLink ID
 Gene
 Chromosomal location
 Function
 Average fold change
Cyst LMP Invasive
4582 MUC1 1q22 Membrane Protein 11.5 20.4 17.8
3725 c-JUN 1p32-p31 Oncogene 7.3 5.1 NC
8714 ABCC3 16q13.1 Drug Transport NC 20.8 11.9
2065 ERBB3 12q13 Cell cycle Regulation NC 11.28 15.7
10788 IQGAP2 5q Cytoskeleton Regulation & Signal Transduction NC 7.4 8.4
595 CCND1 11q13 Cell Cycle Regulation NC 6.1 5.9
10276 NET1 10pl5 Oncogene NC 4.9 7.0
7039 TGF-alpha 2p13 Cell Proliferation NC 4.7 5.6
1894 ECT2 3q26.1 Signal Transduction NC 3.8 6.2
368 ABCC6 17q22 Drug Transport NC 3.4 2.9
3727 JUND 19p13.2 Oncogene NC 1.9 NC
3845 KRAS2 12p12.1 Oncogene NC NC 2.7
99% CDC42 1p36.1 Cytoskeleton Regulation NC NC 2.5
7525 YES1 18p11.31 p11.21 Oncogene NC NC 2.2
1028 CDKN1C 11p15.5 Cell Cycle Inhibition NC –3.6 3.1
10252 SPRV1 4q Cytoskeleton Regulation & Signal Transduction NC –4.5 –3.9
5898 RALA 7p22-p15 Cytoskeleton Regulation & Transduction –2.5 5 4.4
2017 CTIN 11q13 Cytoskeleton Regulation –3.4 NC 4.9
5879 RAC1 7p22 Cytoskeleton Regulation & Signal Transduction –4.4 2.5 NC
857 CAV-1 7p31 Cytoskeleton Regulation & Signal Transduction –20 –10.0 –6.5
7490 WT1 11p13 Tumor Suppressor –6.58 –43.5 –47.6

provides gene annotation information.

 
Microarray validation. To assess the accuracy of the microarray, 14 genes identified as differentially expressed between the mucinous cystadenomas, LMPs, or adenocarcinomas versus normal OSE were evaluated by qRT-PCR (Fig. 4). Although all 14 genes were successfully validated, the quantitative changes exhibited by the selected genes, albeit consistent between the two techniques, did not exactly correlate between the qRT-PCR and microarray analysis. This may be a reflection of the differences in sensitivity between the two techniques. In addition, previous studies have shown that the two techniques are affected by the separation between the PCR primers and microarray probes as well as the number of absent calls made during normalization of the microarray data (27).


Figure 4
Figure 4
View larger version (42K):
[in this window]
[in a new window]
  		Fig. 4. Validation of microarray data using qRT-PCR. Fourteen randomly selected genes were used to validate the microarray data. qRT-PCR data was calculated using the 2–{Delta}{Delta}CT method with mean fold-change values expressed relative to normal OSE (OSE = 1). (A) seven genes identified in all mucinous tumors. Real-time validation confirmed differential expression of (B) seven genes differentially expressed in mucinous LMP tumors and adenocarcinomas, which were not scored as present during normalization. Of these genes, ELF3, ID4, AGR2, NET1, IQGAP2, PROMININ2, and SPROUTY1 have not been previously reported in ovarian cancer.

 
To further validate the microarray results, we selected a differentially expressed gene, AGR2, for which there was an available antibody and did immunohistochemical staining. The AGR2 protein was completely absent in normal ovarian epithelium sections, whereas the mucinous cystadenomas and LMP tumors stained weakly for the protein (Fig. 5A-C). Both grades 1 and 2 adenocarcinomas displayed moderate staining for AGR2 (Fig. 5D).


Figure 5
View larger version (119K):
[in this window]
[in a new window]
  		Fig. 5. Immunohistochemical staining of normal ovarian epithelium (A), mucinous cystadenoma (B), mucinous LMP tumor (C), and mucinous adenocarcinoma (D). Staining intensity score: 1, absent; 2, weak; 3, moderate; 4, strong. All pictures photographed at x400 magnification.

 

   Discussion
Top
 Abstract
 Materials and Methods
 Results
 Discussion
References
 
Mucinous ovarian cancer represents a unique subset of ovarian cancer. These tumors possess a distinct developmental origin and clinical characteristics. There is molecular and pathologic evidence that there is a progression of mucinous cystadenomas to an LMP tumor before developing into an advanced adenocarcinoma. Advanced-stage disease display aggressive, chemoresistant behavior with patients demonstrating a shorter median survival rate (28). This contrasts to other ovarian cancer histologic types, such as papillary serous. Our hierarchical clustering showed that mucinous LMP tumors are clearly distinct from their serous LMP counterparts. There was also a close association among mucinous LMP tumors, mucinous adenocarcinomas, and high-grade serous adenocarcinomas, which may support the more aggressive phenotype associated with mucinous tumors.

In contrast, binary tree analysis showed a clear partition between high-grade serous and the mucinous tumors. This discrepancy may be associated with the unique algorithms underlying each analysis. Whereas hierarchical clustering sequentially merges all of the samples commencing with the first two expression profiles most similar to each other, binary tree prediction begins with all of the specimens comprising each class and splits them according to the division yielding the fewest cross-validated misclassification errors. Therefore, increased heterogeneity in the dendrogram analysis, compared with the binary prediction, is expected particularly when there is a large distance separating the groups. However, association of the mucinous tumors and high-grade serous cancers in the hierarchical clustering tree does suggest the existence of a coregulated subset of gene that may account for their phenotype.

The clustering relationships identified among the mucinous tumors also have important implications for the development of these tumors. Although the origin of mucinous ovarian tumors remains unclear, genetic data suggests that these tumors follow a course of progression commencing with a benign cystadenoma, which is followed by development of an LMP tumor, before the onset of invasive disease. This is evidenced by the frequent occurrence of ras mutations in mucinous LMP tumors and adenocarcinomas (17–20, 28). In contrast, mutations in bRAF and K-ras are observed in serous LMP, but not adenocarcinomas, implying that these tumors may develop independently of each other (29, 30). As shown in the hierarchical clustering analysis, serous LMP tumors and adenocarcinomas group in distinct arms of the dendrogram, whereas mucinous tumors were closely associated in a single branch. Binary tree analysis confirmed this observation with mucinous tumors displaying a high degree of misclassification, compared with their serous counterparts. These data supports a model whereby mucinous ovarian cancer evolves from an LMP intermediary tumor to an adenocarcinoma.

Analyzing mucinous cystadenomas with the LMP and adenocarcinoma samples revealed distinct clustering of the cystadenomas on a separate arm, as would be expected given their unique clinical and molecular behavior (31). Binary tree analysis, however, implied that there is a subset of coregulated genes shared between the cystadenomas and other mucinous tumors. This observation was supported by qRT-PCR validation of AGR2, CAV1, ELF3, ID4, and WT1 in all of the mucinous tumors. Of particular interest were putative markers present among the coexpressed genes, which were not differentially regulated in the serous specimens. Differential expression across all of the tumors suggests that these genes may be associated with the development of the mucinous phenotype. In turn, they may also play a role in the clinical development of mucinous LMP tumors and adenocarcinomas when differentially expressed within these contexts.

As suggested by the hierarchical clustering analysis, a number of genes were also differentially regulated in the LMP tumors and adenocarcinomas but were not differentially regulated in the cystadenomas versus OSE. Included among these genes were NET1 and ERBB3. Previous studies have found that NIH3T3 cells transfected with NET1 show altered growth properties and increased tumorigenicity in vivo, whereas ERBB3 has been shown to promote growth and invasiveness in lung adenocarcinoma (32, 33). Based on these data, up-regulation of NET1 and ERBB3 in LMP and adenocarcinoma tumors may participate in the initiation of the transformation process in mucinous ovarian cancer. Of course, although these genes might be relevant to the biology of the tumor, differential expression of these genes may only be a consequence of tumorigenesis. Further studies need to be carried out to elucidate the mechanisms underlying the actions of these genes in mucinous ovarian cancer.

PathwayAssist software was used to identify coregulated signaling pathways that may contribute to the development of mucinous ovarian cancer. Although a small subset of the genes (MUC1, WT1, CAV1, and c-JUN) contained in the pathway were also differentially expressed in the cystadenomas, most of the genes involved in cell cycling, drug transport, and cytoskeletal rearrangement were only differentially expressed in the LMP tumors and adenocarcinomas. Our data identified ABCC3 and ABCC6, which are involved in the development of multidrug resistance, as up-regulated in both mucinous LMP and mucinous adenocarcinomas compared with mucinous cystadenomas. Previous studies have shown that ABCC3 mRNA levels are elevated in advanced ovarian serous adenocarcinoma (34). ABCC6 expression, however, has not been documented in ovarian carcinomas. Sustained overexpression of ABCC3 and ABCC6 in LMP and advanced mucinous tumors suggests that deregulation of these genes may occur early in the development of the disease. Furthermore, this observation would account for the failure of patients presenting with advanced disease to respond effectively to first-line chemotherapy resulting in poor overall prognosis (9, 34, 35).

A number of genes implicated in cell cycle progression and cellular proliferation were also coexpressed in mucinous LMP and advanced tumors. Transforming growth factor-{alpha}, a strong mitogen (36), and cyclin D1, an important regulator of cell cycle progression (37), were both up-regulated in mucinous LMP tumors and adenocarcinomas. In addition, caveolin-1 (CAV-1) and sprouty1 (SPRY1), which are negative regulators of cell cycle progression, were down-regulated in the LMP and adenocarcinomas (38, 39).

Interestingly, genes whose proteins modulate cell morphology were up-regulated in mucinous adenocarcinoma and not mucinous LMP specimens. These included CDC42, ECT2, IQGAP2, and Cortactin. Cross-talk among these genes activates various mechanisms involved in cytoskeletal rearrangement (40, 41). Although there are no reports on the expression of these genes in ovarian cancers, studies in other neoplasia and cancer cell lines show that CDC42 is overexpressed in breast tumor specimens (42), ECT2 in HeLa cells (43), and Cortactin in breast cancer, where its expression correlates with poor prognosis (44). The expression of IQGAP2 in cancer is poorly documented; however, IQGAP1 is up-regulated in colorectal (45) and gastric cancer (46). Up-regulation of these genes in mucinous adenocarcinomas, rather than LMP tumors, suggests that they may contribute to the aggressive nature of mucinous ovarian cancer as the disease progresses to an advanced state.

In summary, hierarchical clustering revealed that mucinous LMP tumors are distinct from their serous counterparts; however, there was a close molecular association between the mucinous LMP tumors, mucinous adenocarcinomas, and serous adenocarcinomas. Further analysis of the relationships among these tumors confirmed the close association of mucinous LMP tumors and adenocarcinomas in agreement with current molecular data. Our data also suggests that mucinous cystadenomas may share common molecular features with malignant mucinous ovarian adenocarcinoma. Finally, pathway analysis identified gene interactions that may influence mucinous ovarian tumorigenesis, as well as genes whose functions might mediate the phenotypes typically associated with these tumors. The data generated from this study form an important basis for further investigation into the relevance of these genes in the efficacious management and prognostic assessment of mucinous ovarian cancer.


   Footnotes
 
Grant support: Dana-Faber/Harvard Ovarian Cancer Specialized Programs of Research Excellence grant P50CA165009 and R33CA103595 from the NIH, Department of Health and Human Services, and by the Intramural Research Program of the NIH, National Cancer Institute.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Note: Supplementary data for this article are available at Clinical Cancer Research Online (http://clincancerres.aacrjournals.org/).

Received 5/24/05; revised 8/31/05; accepted 11/17/05.


   References
Top
 Abstract
 Materials and Methods
 Results
 Discussion
 References
 

  1. Jemal A, Murray T, Ward E, et al. Cancer statistics, 2005. CA Cancer J Clin 2005;55:10–30.[Abstract/Free Full Text]
  2. Verheijen RH, von Mensdorff-Pouilly S, van Kamp GJ, Kenemans P. CA 125: fundamental and clinical aspects. Semin Cancer Biol 1999;9:117–24.[CrossRef][Medline]
  3. Bast RC, Jr., Urban N, Shridhar V, et al. Detection of ovarian cancer: promise and reality. In: Stack MS, Fishman DA, editors. Cancer treatment and research. Vol 107. Boston: Kluwer Academic; 2002. p. 61–97.
  4. Scully RE, Young RH, Clement PB. Tumors of the ovary, maldeveloped gonads, fallopian tubes and broad ligament. Vol 3, fascicle 23. Washington (District of Columbia): Armed Forces Institute of Pathology; 1998.
  5. Hart WR. Mucinous tumors of the ovary: a review. Int J Gynecol Pathol 2005;24:4–25.[Medline]
  6. Creasman WT, Park R, Norris H, Disaia PJ, Morrow CP, Hreshchyshyn MM. Stage I borderline ovarian tumors. Obstet Gynecol 1982;59:93–6.[Abstract/Free Full Text]
  7. Casey AC, Bell DA, Lage JM, Fuller AF, Nikrui N, Rice LW. Epithelial ovarian tumors of borderline malignancy: long-term follow-up. Gynecol Oncol 1993;50:316–22.[CrossRef][Medline]
  8. Trope C, Kaern J, Vergote IB, Kristensen G, Abeler V. Are borderline tumors of the ovary overtreated both surgically and systemically? A review of four prospective randomized trials including 253 patients with borderline tumors. Gynecol Oncol 1993;51:236–43.[CrossRef][Medline]
  9. Hess V, A'Hern R, Nasiri N, et al. Mucinous epithelial ovarian cancer: a separate entity requiring specific treatment. J Clin Oncol 2004;22:1040–4.[Abstract/Free Full Text]
  10. Omura GA, Brady MF, Homesley HD, et al. Long-term follow-up and prognostic factor analysis in advanced ovarian carcinoma: the Gynecologic Oncology Group experience. J Clin Oncol 1991;9:1138–50.[Abstract]
  11. Kosary CL. FIGO stage, histology, histologic grade, age and race as prognostic factors in determining survival for cancers of the female gynecological system: an analysis of 1973-87 SEER cases of cancers of the endometrium, cervix, ovary, vulva, and vagina. Semin Surg Oncol 1994;10:31–46.[Medline]
  12. Makar AP, Baekelandt M, Trope CG, Kristensen GB. The prognostic significance of residual disease, FIGO substage, tumor histology, and grade in patients with FIGO stage III ovarian cancer. Gynecol Oncol 1995;56:175–80.[CrossRef][Medline]
  13. Stenback F. Benign, borderline and malignant serous cystadenomas of the ovary. A transmission and scanning electron microscopical study. Pathol Res Pract 1981;172:58–72.[Medline]
  14. Puls LE, Powell DE, DePriest PD, et al. Transition from benign to malignant epithelium in mucinous and serous ovarian cystadenocarcinoma. Gynecol Oncol 1992;47:53–7.[CrossRef][Medline]
  15. Scully RE, Bell DA, Abu-Jawdeh GM. Update on early ovarian cancer and cancer developing in benign ovarian tumours. London: Chapman and Hall Medical; 1995.
  16. Garrett AP, Lee KR, Colitti CR, Muto MG, Berkowitz RS, Mok SC. K-ras mutation may be an early event in mucinous ovarian tumorigenesis. Int J Gynecol Pathol 2001;20:244–51.[Medline]
  17. Mok SC, Bell DA, Knapp RC, et al. Mutation of K-ras protooncogene in human ovarian epithelial tumors of borderline malignancy. Cancer Res 1993;53:1489–92.[Abstract/Free Full Text]
  18. Ichikawa Y, Nishida M, Suzuki H, et al. Mutation of K-ras protooncogene is associated with histological subtypes in human mucinous ovarian tumors. Cancer Res 1994;54:33–5.[Abstract/Free Full Text]
  19. Cuatrecasas M, Villanueva A, Matias-Guiu X, Prat J. K-ras mutations in mucinous ovarian tumors: a clinicopathologic and molecular study of 95 cases. Cancer 1997;79:1581–6.[CrossRef][Medline]
  20. Pieretti M, Hopenhayn-Rich C, Khattar NH, Cao Y, Huang B, Tucker TC. Heterogeneity of ovarian cancer: relationships among histological group, stage of disease, tumor markers, patient characteristics, and survival. Cancer Invest 2002;20:11–23.[CrossRef][Medline]
  21. Donninger H, Bonome T, Radonovich M, et al. Whole genome expression profiling of advance stage papillary serous ovarian cancer reveals activated pathways. Oncogene 2004;23:8065–77.[CrossRef][Medline]
  22. Bonome T, Lee JY, Park DC, et al. Expression profiling of serous low malignant potential, low-grade, and high-grade tumors of the ovary. Cancer Res 2005;65:10602–12.[Abstract/Free Full Text]
  23. Zorn KK, Jazaeri AA, Awtrey CS, et al. Choice of normal ovarian control influences determination of differentially expressed genes in ovarian cancer expression profiling studies. Clin Cancer Res 2003;9:4811–8.[Abstract/Free Full Text]
  24. Li C, Wong WH. Model-based analysis of oligonucleotide arrays: expression index computation and outlier detection. PNAS 2001;98:31–6.[Abstract/Free Full Text]
  25. Livak KJ, Schmittgen TD. Analysis of relative gene expression data using real-time quantitative PCR and the 2[–{Delta}{Delta}C(T)] method. Methods 2001;25:402–8.[CrossRef][Medline]
  26. Zhang JS, Gong A, Cheville JC, Smith DI, Young CY. AGR2, an androgen-inducible secretory protein overexpressed in prostate cancer. Genes Chromosomes Cancer 2005;43:249–59.[CrossRef][Medline]
  27. Etienne W, Meyer MH, Peppers J, Meyer RA, Jr. Comparison of mRNA gene expression by RT-PCR and DNA microarray. Biotechniques 2004;36:618–20, 22, 24–6.[Medline]
  28. Tamakoshi K, Kikkawa F, Nakashima N, et al. Clinical behavior of borderline ovarian tumors: a study of 150 cases. J Surg Oncol 1997;64:147–52.[Medline]
  29. Teneriello MG, Ebina M, Linnoila RI, et al. p53 and Ki-ras gene mutations in epithelial ovarian neoplasms. Cancer Res 1993;53:3103–8.[Abstract/Free Full Text]
  30. Shih IM, Kurman RJ. Ovarian tumorigenesis: a proposed model based on morphological and molecular genetic analysis. Am J Pathol 2004;164:1511–8.[Abstract/Free Full Text]
  31. Lee KR, Scully RE. Mucinous tumors of the ovary: a clinicopathologic study of 196 borderline tumors (of intestinal type) and carcinomas, including an evaluation of 11 cases with "pseudomyxoma peritonei". Am J Surg Pathol 2000;24:1447–64.[Medline]
  32. Chan AM, Takai S, Yamada K, Miki T. Isolation of a novel oncogene, NET1, from neuroepithelioma cells by expression cDNA cloning. Oncogene 1996;12:1259–66.[Medline]
  33. Sithanandam G, Fornwald LW, Fields J, Anderson LM. Inactivation of ErbB3 by siRNA promotes apoptosis and attenuates growth and invasiveness of human lung adenocarcinoma cell line A549. Oncogene 2005;24:1847–59.[CrossRef][Medline]
  34. Ohishi Y, Oda Y, Uchiumi T, et al. ATP-binding cassette superfamily transporter gene expression in human primary ovarian carcinoma. Clin Cancer Res 2002;8:3767–75.[Abstract/Free Full Text]
  35. McGuire V, Jesser CA, Whittemore AS. Survival among U.S. women with invasive epithelial ovarian cancer. Gynecol Oncol 2002;84:399–403.[CrossRef][Medline]
  36. Lee DC, Fenton SE, Berkowitz EA, Hissong MA. Transforming growth factor {alpha}: expression, regulation, and biological activities. Pharmacol Rev 1995;47:51–85.[Medline]
  37. Baldin V, Lukas J, Marcote MJ, Pagano M, Draetta G. Cyclin D1 is a nuclear protein required for cell cycle progression in G1. Genes Dev 1993;7:812–21.[Abstract/Free Full Text]
  38. Williams TM, Lisanti MP. Caveolin-1 in oncogenic transformation, cancer, and metastasis. Am J Physiol Cell Physiol 2005;288:C494–506.[Abstract/Free Full Text]
  39. Gross I, Bassit B, Benezra M, Licht JD. Mammalian sprouty proteins inhibit cell growth and differentiation by preventing ras activation. J Biol Chem 2001;276:46460–8.[Abstract/Free Full Text]
  40. Hall A. Rho GTPases and the actin cytoskeleton. Science 1998;279:509–14.[Abstract/Free Full Text]
  41. Sahai E, Marshall CJ. RHO-GTPases and cancer. Nat Rev Cancer 2002;2:133–42.[CrossRef][Medline]
  42. Fritz G, Just I, Kaina B. Rho GTPases are over-expressed in human tumors. Int J Cancer 1999;81:682–7.[CrossRef][Medline]
  43. Tatsumoto T, Xie X, Blumenthal R, Okamoto I, Miki T. Human ECT2 is an exchange factor for Rho GTPases, phosphorylated in G2/M phases, and involved in cytokinesis. J Cell Biol 1999;147:921–8.[Abstract/Free Full Text]
  44. Hui R, Ball JR, Macmillan RD, et al. EMS1 gene expression in primary breast cancer: relationship to cyclin D1 and oestrogen receptor expression and patient survival. Oncogene 1998;17:1053–9.[CrossRef][Medline]
  45. Nabeshima K, Shimao Y, Inoue T, Koono M. Immunohistochemical analysis of IQGAP1 expression in human colorectal carcinomas: its overexpression in carcinomas and association with invasion fronts. Cancer Lett 2002;176:101–9.[CrossRef][Medline]
  46. Sugimoto N, Imoto I, Fukuda Y, et al. IQGAP1, a negative regulator of cell-cell adhesion, is upregulated by gene amplification at 15q26 in gastric cancer cell lines HSC39 and 40A. J Hum Genet 2001;46:21–5.[CrossRef][Medline]



This article has been cited by other articles:


Home page
 JCOHome page
C. N. Landen Jr, M. J. Birrer, and A. K. Sood
Early Events in the Pathogenesis of Epithelial Ovarian Cancer
J. Clin. Oncol., February 20, 2008; 26(6): 995 - 1005.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplementary Data
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Wamunyokoli, F. W.
Right arrow Articles by Birrer, M. J.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Wamunyokoli, F. W.
Right arrow Articles by Birrer, M. J.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Cancer Research Clinical Cancer Research
Cancer Epidemiology Biomarkers & Prevention Molecular Cancer Therapeutics
Molecular Cancer Research Cancer Prevention Research
Cancer Prevention Journals Portal Cancer Reviews Online
Annual Meeting Education Book Cell Growth & Differentiation